Occidental said:
My thanks to everyone who responded.
The answer appears to be that it will only take the original Phenom
9550, which is no longer sold. There are "used" models of this unit
available on ebay. However, they appear to be sold without heatsink.
My CPU is glued to the heatsink. Can it be removed and the new one
"attached"?
Here is what happened: Back when the PC was functioning, I noticed
that from time to time (usually during a comp-intensive process
(winzip, chess program)) the case cooling fan would temporarily go
into high-rev mode. Two days ago, the PC shut down while in operation.
The screen was blue (not BSOD, no text, just blue). Reboot did not
work. With the case open I discovered heavy dust accumulation all over
the CPU heatsink, including within the cooling fins. I cleaned it up
and tried to reboot. I got some of the way through the boot process,
stopping at the "insert bootable media" message. Rebooted again and
didn't get that far. Rebooted again and got nothing at all. Rebooted
this morning and got nothing, then auto shutdown, auto restart, auto
shutdown etc, which had not happened before. Does that sound like a
dead CPU? My only reservation is that I would expect an overheating
CPU to send a message to the monitor and then shut itself down. This
should be easy to implement by the OEM. Even a hair dryer has a
thermal cutout.
There is actually a material, that can be used for gluing integrated
circuits to heatsinks. It's called "Thermal Epoxy", due to its ability
to conduct heat.
But that would not be a preferred material, because it interferes with
the need to make repairs to the motherboard. Even when the motherboards
are new, chips need to be changed out and so on. It allows motherboard
yield at the plant to be 99.7% (only 0.3% thrown away, which is still
a large amount).
Instead, thermal paste or phase change material, is placed between the
heatsink and CPU. You can clean the material off, and apply fresh paste.
See "how to apply thermal compound", in the right hand column.
http://www.arcticsilver.com/cmq2.html
You can try to clean off the residue with IPA (isopropyl alcohol) but
it's not really an idea solvent. The nice thing about IPA is it won't
damage anything on the motherboard. You can also get purpose-made
solvents, in kit form.
http://www.newegg.com/Product/Product.aspx?Item=N82E16835100010
The main thing is, try not to get paste into the electrical area of
the processor, like on the pins of the processor or the contact area
in the socket. Some pastes are slightly capacitive, and can affect
signal shape if they ooze over everything.
To remove a CPU, try to heat it up first, by using the computer. Then,
shut down the computer, turn off the power, and while the heatsink is
mildly warm, remove the clamp (pins or bar or whatever is used), then
give the heatsink a slight twist to loosen the grip of the paste.
*******
If you had a thermal problem, it could be as simple as a broken clamp
on the heatsink. Some AMD motherboards, the heatsink downforce is set too
high, and the bar that holds the heatsink, stresses the plastic tab
on the socket until the plastic snaps. So you'd start by inspecting
to see whether the clamping method is still secure.
"Insert bootable media" means the hard drive was not detected. Note
that, if you press the reset button on the PC, that is guaranteed to
reset an IDE (ribbon cable) connected disk. But SATA drives, the reset
button doesn't do anything of note. To guarantee a SATA drive is reset,
you have to power off the PC, wait 20 seconds, then power on again.
(Note - do not "toggle" the power switch on the back in rapid succession.
Allow the inrush limiter to cool off, after the PC is switched off.
Allow a measurable time, before powering on again. That's to avoid
blowing out the supply, or causing a supply protection feature from
malfunctioning.)
The "Insert bootable media" could mean the hard drive was "hung" and
had stopped responding to commands. That could mean the disk is about
to fail soon.
I've had one "SATA hang" here, there were no long term consequences, and
I continue to use the drive. So these things happen, but should be a
relatively rare occurrence.
You can get disk test software from the disk manufacturer. For example, if
the brand on the label of the hard drive said "Seagate", I could get
Seatools for Windows or Seatools for DOS from the seagate.com web site.
It will test the hard drive, look at SMART statistics, and report
whether the drive has completely failed or not. Other third-party
programs can read the SMART statistics, such as HDTune from hdtune.com
(free edition 2.55). To do that now, you'd need to move the suspect
hard drive, into a good computer, so there is a working setup to do the
testing with.
*******
An actual "THERMTRIP" event, where the CPU overheats, causes the PC
to shut off, by using the PS_ON# control signal. The PC will not come on
again, until you turn off the power at the back, wait 20 seconds, and
turn it on again. If the CPU is still blazing hot, it would turn off
almost immediately. For example, if the CPU clamp was broken, and the
CPU heatsink was hanging there, THERMTRIP could trigger in a couple
seconds to protect the CPU. THERMTRIP is based on a temperature measurement
done on the CPU silicon die, so it's "right next to the heat".
I don't see anything to suggest the problem is CPU. There are other
things that could cause symptoms. Bad RAM. A motherboard problem.
A bad hard drive.
If I was working on it, I would have to "simplify" the setup, and
test a bit of it at a time. That means disconnecting things, then
look at symptoms. For example, set up motherboard, CPU, CPU heatsink,
no RAM, no optical drive connected, no hard drive connected, and
see if the power remains on when the motherboard is powered up.
If there is a THERMTRIP problem, presumably it is still going to be
present, when the reduced set of components are used. With no RAM installed,
the BIOS POST code will be in a "beep loop" and that is a simple
computer program of sorts. So as long as the beeping continues, the
CPU is running the BIOS code (while no RAM sticks are present).
You can dream up test cases like that, where the computer is simplified
and (relatively) non-functional, where you can still get symptom information
from the setup.
For some possibilities, while you can make measurements, component swap
is faster. Like, you can't really completely quantify an ATX power supply,
with things like a multimeter - it doesn't give a complete health analysis.
So it may be faster to just place a replacement supply in the machine
temporarily, to see if the symptoms change. Not all of the debugging
techniques need use "science" to get an answer. To properly test a
supply, takes a "Chroma Tester", a rack of equipment six feet high,
and it does a fully automated test of ATX supply parameters (static and
dynamic). And it could probably find a "weakened" supply for you. To do
that at home, yes you could do it, but it would be a month long project
with trips to the electronics store to construct dummy loads and the
like, to do even a fraction of the tests. For most users, it's easier
to just swap in that spare supply they keep for such occasions. Supply
failure is so frequent, that for some repair people, it might be their
first test. If you buy $20 ATX supplies, you might see them fail
four times a year (as reported on a Newegg reviewer comment for
a cheap supply they were buying - they said they'd bought four
of them sequentially, and they were happy that the supplies were
so cheap - a form of "false economics", as an $80 supply probably
would have lasted five years).
Staying in a loop doing restarts, could be bad RAM (CPU goes crazy
because it's running non-code), could be a weak supply (+5VSB drops
too low, causing motherboard to reset), but isn't likely to be a bad
CPU. If you saw even one proper BSOD while the computer was running,
and a particular obscure error message, that might have given a
hint of a damaged processor. But I'm not seeing anything specific
enough in your report, to narrow it down to one component. Based on
what you've reported so far, it's not definitely the CPU. Especially
"insert bootable media" - to print that on the screen, the CPU was
still running quite nicely at the time.
So you'll have to methodically test, and cook up some reduced test
cases, to prove the components in the machine. And since you got into
a reboot loop, try another ATX supply (could be weakened outputs). After
you've removed a few supplies, keeping track of the wires is no problem
at all. I can tell you though, the first one I did, I very carefully
labeled everything, so I'd get it back together again
The first one
is scary if you've never done one before ...
Paul
